def _test(res=''):
raw_data, labels = readfile("../../data/UCI/" + res + ".csv")
metric = "d2h"
final = {}
final_auc = {}
e_value = [0.05, 0.1, 0.2]
start_time = time.time()
dic = {}
dic_func = {}
for mn in range(500 + file_inc[res] * 10, 521 + file_inc[res] * 10):
for e in e_value:
np.random.seed(mn)
seed(mn)
preprocess = [
standard_scaler, minmax_scaler,
maxabs_scaler, [robust_scaler] * 20, kernel_centerer,
[quantile_transform] * 200, normalizer, [binarize] * 100
] # ,[polynomial]*5
MLs = [NB, [KNN] * 20, [RF] * 50, [DT] * 30,
[LR] * 50] # [SVM]*100,
preprocess_list = unpack(preprocess)
MLs_list = unpack(MLs)
combine = [[r[0], r[1]]
for r in product(preprocess_list, MLs_list)]
if e not in final_auc.keys():
final_auc[e] = []
dic[e] = {}
func_str_dic = {}
func_str_counter_dic = {}
lis_value = []
dic_auc = {}
for i in combine:
scaler, tmp1 = i[0]()
model, tmp2 = i[1]()
string1 = tmp1 + "|" + tmp2
func_str_dic[string1] = [scaler, model]
func_str_counter_dic[string1] = 0
counter = 0
while counter != 200:
if counter not in dic_func.keys():
dic_func[counter] = []
keys = [k for k, v in func_str_counter_dic.items() if v == 0]
key = _randchoice(keys)
cut_off = int(len(raw_data) * 0.8)
scaler, model = func_str_dic[key]
df1 = transform(raw_data, scaler)
df1.loc[:, "bug"] = labels
train_data, test_data = df1.iloc[:cut_off, :], df1.iloc[
cut_off:, :]
measurement = run_model(train_data,
test_data,
model,
metric,
training=-1)
if all(abs(t - measurement) > e for t in lis_value):
lis_value.append(measurement)
func_str_counter_dic[key] += 1
else:
func_str_counter_dic[key] += -1
if counter not in dic[e].keys():
dic[e][counter] = []
dic_func[counter] = []
if e == 0.05:
dic_func[counter].append(key)
dic[e][counter].append(min(lis_value))
dic_auc[counter] = min(lis_value)
counter += 1
dic1 = OrderedDict(sorted(dic_auc.items(),
key=itemgetter(0))).values()
area_under_curve = round(auc(list(range(len(dic1))), dic1), 3)
final[e] = dic_auc
final_auc[e].append(area_under_curve)
total_run = time.time() - start_time
final_auc["temp"] = final
final_auc["time"] = total_run
final_auc["counter_full"] = dic
final_auc["settings"] = dic_func
print(final_auc)
with open('dump/d2h_' + res + '.pickle', 'wb') as handle:
pickle.dump(final_auc, handle)
def _test(res=''):
raw_data, labels = readfile("../../data/textmining/" + res + ".txt")
metric = "Pf_Auc"
final = {}
final_auc = {}
e_value = [0.2]
start_time = time.time()
dic = {}
dic_func = {}
for mn in range(500 + file_inc[res] * 10, 511 + file_inc[res] * 10):
for e in e_value:
transformation = [[TF] * 30, [TFIDF] * 30, [HASHING] * 8,
[LDA_] * 50]
preprocess = [no_transformation]
MLs = [NB, [KNN] * 20, [RF] * 50, [DT] * 30, [LR] * 50]
preprocess_list = unpack(preprocess)
MLs_list = unpack(MLs)
trans_list = unpack(transformation)
combine = [[r[0], r[1], r[2]]
for r in product(trans_list, preprocess_list, MLs_list)]
if e not in final_auc.keys():
final_auc[e] = []
dic[e] = {}
func_str_dic = {}
func_str_counter_dic = {}
lis_value = []
dic_auc = {}
for i in combine:
trans, tmp = i[0]()
scaler, tmp1 = i[1]()
model, tmp2 = i[2]()
string1 = tmp + "|" + tmp1 + "|" + tmp2
func_str_dic[string1] = [trans, scaler, model]
func_str_counter_dic[string1] = 0
counter = 0
while counter != 40:
print(counter)
if counter not in dic_func.keys():
dic_func[counter] = []
keys = [k for k, v in func_str_counter_dic.items() if v == 0]
key = _randchoice(keys)
cut_off = int(len(raw_data) * 0.8)
vector, scaler, model = func_str_dic[key]
df = extraction(raw_data, vector)
df1 = transform(df, scaler)
df1.loc[:, "bug"] = labels
train_data, test_data = df1.iloc[:cut_off, :], df1.iloc[
cut_off:, :]
measurement = run_model(train_data,
test_data,
model,
metric,
training=-1)
if all(abs(t - measurement) > e for t in lis_value):
lis_value.append(measurement)
func_str_counter_dic[key] += 1
else:
func_str_counter_dic[key] += -1
if counter not in dic[e].keys():
dic[e][counter] = []
dic_func[counter] = []
dic_func[counter].append(key)
dic[e][counter].append(min(lis_value))
dic_auc[counter] = min(lis_value)
counter += 1
dic1 = OrderedDict(sorted(dic_auc.items(),
key=itemgetter(0))).values()
area_under_curve = round(auc(list(range(len(dic1))), list(dic1)),
3)
final[e] = dic_auc
final_auc[e].append(area_under_curve)
total_run = time.time() - start_time
final_auc["temp"] = final
final_auc["time"] = total_run
final_auc["counter_full"] = dic
final_auc["settings"] = dic_func
print(final_auc)
def _test(res=''):
#df=readfile("../data/defect/"+res+".csv")
paths = [os.path.join(data_path, file_name) for file_name in file_dic[res]]
train_df = pd.concat([pd.read_csv(path) for path in paths[:-1]], ignore_index=True)
test_df = pd.read_csv(paths[-1])
train_df, test_df = train_df.iloc[:, 3:], test_df.iloc[:, 3:]
train_size=train_df["bug"].count()
df=pd.concat([train_df,test_df],ignore_index=True)
df['bug']=df['bug'].apply(lambda x: 0 if x == 0 else 1)
metric="d2h"
final = {}
final_auc={}
e_value = [0.025, 0.05, 0.1, 0.2]
start_time=time.time()
dic={}
dic_func={}
for mn in range(500+file_inc[res]*10,521+file_inc[res]*10):
for e in e_value:
np.random.seed(mn)
seed(mn)
preprocess = [standard_scaler, minmax_scaler, maxabs_scaler, [robust_scaler] * 20, kernel_centerer,
[quantile_transform] * 200
, normalizer, [binarize] * 100] # ,[polynomial]*5
MLs = [NB, [KNN] * 20, [RF] * 50, [DT] * 30, [LR] * 50] # [SVM]*100,
preprocess_list = unpack(preprocess)
MLs_list = unpack(MLs)
combine = [[r[0], r[1]] for r in product(preprocess_list, MLs_list)]
if e not in final_auc.keys():
final_auc[e]=[]
dic[e] = {}
func_str_dic = {}
func_str_counter_dic = {}
lis_value = []
dic_auc={}
for i in combine:
scaler, tmp1 = i[0]()
model, tmp2 = i[1]()
string1 = tmp1 + "|" + tmp2
func_str_dic[string1] = [scaler, model]
func_str_counter_dic[string1] = 0
counter=0
while counter!=1000:
if counter not in dic_func.keys():
dic_func[counter]=[]
try:
keys = [k for k, v in func_str_counter_dic.items() if v == 0]
key = _randchoice(keys)
scaler,model=func_str_dic[key]
df1=transform(df,scaler)
train_data, test_data = df1.iloc[:train_size,:], df1.iloc[train_size:,:]
measurement = run_model(train_data, test_data, model, metric,training=-2)
if all(abs(t - measurement) > e for t in lis_value):
lis_value.append(measurement)
func_str_counter_dic[key] += 1
else:
func_str_counter_dic[key] += -1
if counter not in dic[e].keys():
dic[e][counter] = []
dic_func[counter]=[]
if e == 0.025:
dic_func[counter].append(key)
dic[e][counter].append(min(lis_value))
dic_auc[counter]=min(lis_value)
# sss = StratifiedShuffleSplit(n_splits=1, test_size=0.2)
# data,labels=df1[df1.columns[:-2]], df1[df1.columns[-2]]
# for train_index, test_index in sss.split(data,labels):
# train_data, test_data=df1.iloc[train_index], df1.iloc[test_index]
# measurement=run_model(train_data,test_data,model,metric)
#
# if all(abs(t-measurement)>e for t in lis_value):
# lis_value.append(measurement)
# func_str_counter_dic[key] += 1
# else:
# func_str_counter_dic[key] += -1
#
# dic[counter]=max(lis_value)
counter+=1
except:
pass
dic1 = OrderedDict(sorted(dic_auc.items(), key=itemgetter(0))).values()
area_under_curve=round(auc(list(range(len(dic1))), dic1), 3)
final[e]=dic_auc
final_auc[e].append(area_under_curve)
total_run=time.time()-start_time
final_auc["temp"]=final
final_auc["time"] = total_run
final_auc["counter_full"]=dic
final_auc["settings"]=dic_func
print(final_auc)
with open('../../dump/d2h_' + res + '.pickle', 'wb') as handle:
pickle.dump(final_auc, handle)
def _test(res=''):
paths = [os.path.join(data_path, file_name) for file_name in file_dic[res]]
train_df = pd.concat([pd.read_csv(path) for path in paths[:-1]],
ignore_index=True)
test_df = pd.read_csv(paths[-1])
train_df, test_df = train_df.iloc[:, 3:], test_df.iloc[:, 3:]
train_size = train_df["bug"].count()
df = pd.concat([train_df, test_df], ignore_index=True)
df['bug'] = df['bug'].apply(lambda x: 0 if x == 0 else 1)
metric = "popt20"
final = {}
final_auc = {}
e_value = [0.2]
start_time = time.time()
dic = {}
dic_func = {}
for mn in range(500 + file_inc[res] * 10, 520 + file_inc[res] * 10):
for e in e_value:
preprocess = [standard_scaler, minmax_scaler,
[normalizer] * 10] # ,[polynomial]*5
MLs = [[DeepLearner] * 20] # [SVM]*100,
preprocess_list = unpack(preprocess)
MLs_list = unpack(MLs)
combine = [[r[0], r[1]]
for r in product(preprocess_list, MLs_list)]
if e not in final_auc.keys():
final_auc[e] = []
dic[e] = {}
func_str_dic = {}
func_str_counter_dic = {}
lis_value = []
dic_auc = {}
for i in combine:
scaler, tmp1 = i[0]()
model, tmp2 = i[1]()
string1 = tmp1 + "|" + tmp2
func_str_dic[string1] = [scaler, model]
func_str_counter_dic[string1] = 0
counter = 0
while counter != 30:
if counter not in dic_func.keys():
dic_func[counter] = []
try:
keys = [
k for k, v in func_str_counter_dic.items() if v == 0
]
key = _randchoice(keys)
scaler, model = func_str_dic[key]
df1 = transform(df, scaler)
train_data, test_data = df1.iloc[:train_size, :], df1.iloc[
train_size:, :]
measurement = run_model(train_data,
test_data,
model,
metric,
training=-2)
if all(abs(t - measurement) > e for t in lis_value):
lis_value.append(measurement)
func_str_counter_dic[key] += 1
else:
func_str_counter_dic[key] += -1
if counter not in dic[e].keys():
dic[e][counter] = []
dic_func[counter] = []
if e == 0.05:
dic_func[counter].append(key)
dic[e][counter].append(max(lis_value))
dic_auc[counter] = max(lis_value)
counter += 1
except:
raise
dic1 = OrderedDict(sorted(dic_auc.items(),
key=itemgetter(0))).values()
area_under_curve = round(auc(list(range(len(dic1))), list(dic1)),
3)
print("AUC: ", area_under_curve)
final[e] = dic_auc
final_auc[e].append(area_under_curve)
total_run = time.time() - start_time
final_auc["temp"] = final
final_auc["time"] = total_run
final_auc["counter_full"] = dic
final_auc["settings"] = dic_func
print(final_auc)